Warning: mkdir(): Permission denied in /home/virtual/lib/view_data.php on line 81

Warning: fopen(upload/ip_log/ip_log_2024-03.txt): failed to open stream: No such file or directory in /home/virtual/lib/view_data.php on line 83

Warning: fwrite() expects parameter 1 to be resource, boolean given in /home/virtual/lib/view_data.php on line 84
Study on Creating A Classifier for Grading of Bladder Carcinoma Based on Computerized Method.
Skip Navigation
Skip to contents

J Pathol Transl Med : Journal of Pathology and Translational Medicine

OPEN ACCESS
SEARCH
Search

Articles

Page Path
HOME > J Pathol Transl Med > Volume 36(3); 2002 > Article
Original Article Study on Creating A Classifier for Grading of Bladder Carcinoma Based on Computerized Method.
Hyun Ju Choi, Hye Kyoung Yoon, Heung Kook Choi
Journal of Pathology and Translational Medicine 2002;36(3):154-162
DOI: https://doi.org/
1Medical Image Technology Laboratory, School of Information and Computer Engineering, Inje University, Gimhae, Korea. hjchoi@mitl.inje.ac.kr
2Department of Pathology, Busan Paik Hospital, College of Medicine, Inje University, Busan, Korea.
  • 1,396 Views
  • 13 Download
  • 0 Crossref
  • 0 Scopus

BACKGROUND
We have described an objective and reproducible classification method for grading malignancy in the Feulgen stained bladder carcinoma. To create an optimized classifier for malignancy grading of histological bladder carcinoma cell images, it is necessary to extract the features that accurately describle the order/disorder of the nuclear variation and to evaluate the significance of the features. Above all, features selection considered about the correlation of features is very important, because the performance of the classification method depends on the selected features.
METHODS
First, we acquired 40 representative histological bladder carcinoma cell images from each of four groups (Grade 1, Grade 2A, Grade 2B, Grade 3) and extracted morphology features, texture features and the texture features of wavelet transformed images. Second, we evaluated the significance of the extracted features using variance analysis. Third, we created classifiers for each selected feature and its combination set using discriminant analysis. Finally, we compared and analyzed the correct classification rate of each classifer.
RESULTS
The optimized classifier was created from the combination of morphology features, texture features and the texture features of wavelet transformed images.
CONCLUSIONS
We found that the correlation of features is more important than one feature's great significance in grading the malignancy of bladder carcinoma, and we have confirmed that the correct classification rate is determined by feature extractin, feature evaluation and feature selection.

Related articles

J Pathol Transl Med : Journal of Pathology and Translational Medicine